Amazon keyword research for PPC is the process of finding the search terms worth bidding on, then letting real ad data decide which ones keep their budget. The most reliable source is your own search term report, because it ties each search to what your ads spent and the orders that followed. Brand Analytics, Amazon's search bar, competitor listings, and keyword tools are best used to nominate candidates the report then judges.
Most pages ranking for this topic come from keyword-tool vendors, so they start with the tool. I start with the account. A pattern I run into often in audits: a keyword list built once from a tool export at launch, next to a search term report nobody has read since. This guide ranks the five sources I use by how far I trust them, shows why search volume is the wrong number to sort by, walks through my process, and prices a keyword list on real numbers before you pay for it.
Where do Amazon PPC keywords come from? Five sources, ranked
Amazon PPC keywords come from five places: your own search term report and auto campaign harvest, Brand Analytics search query data, Amazon's search bar suggestions, competitor listings and ASINs, and third-party keyword tools. I rank them by one test: how close the data sits to a shopper buying your product. Your own report is closest; tool estimates sit furthest away.
| Source | What it gives you | Its weakness |
|---|---|---|
| 1. Your search term report and auto campaign harvest | Searches that got your ad clicked, with the spend, orders, and sales each produced for your listing. Auto campaigns add searches and ASINs you would not have tried. | Blind to any search your ads never entered or never got clicked on. |
| 2. Brand Analytics search query data (brand-registered sellers) | Amazon's own numbers for searches where your brand appears: market-wide clicks and purchases next to your share. | Brand Registry only. Blends ad and organic traffic, shows no ad cost, and runs thin for a young brand. |
| 3. Amazon's search bar suggestions | Real shopper phrasing, free: the modifiers, sizes, and uses people attach to your root term. | No volume, clicks, or conversions. It shows people type a phrase, not how many or whether they buy. |
| 4. Competitor listings and ASINs | The vocabulary already selling in your niche, plus ASINs you can target directly. | Shows what a competitor chose to write, not what converts. Their reviews and price may be doing the work. |
| 5. Third-party keyword tools | Long candidate lists fast, estimated search volumes, and estimates of which searches a competitor shows up for. | Estimates that often disagree between tools, and none can see your conversion rate. |
Why your own search term report beats any keyword tool
A keyword tool estimates how many people search a phrase. Your search term report shows what happened when shoppers typed it and clicked your ad: what the clicks cost and whether orders followed. Profit is set by what a click costs and what it earns, and search volume appears nowhere in that math. Only the report measures both.
Reduce ACOS to one click and the reason is plain: it is what the click cost divided by what it brought back, and what it brought back is conversion rate times order value. Every input belongs to your listing, your price, or your auction. Volume only decides how much money flows through a keyword, not whether it returns. Amazon describes the search term report as the searches that got your ad clicked at least once, which makes it the only source here that pairs each search with your spend and your orders.
The clearest example I can show you is on my results page. A $45K/mo skincare brand was bleeding spend on variant search terms that never converted, with wasted spend at 41%. My fix was to drop the long tail, put the spend behind the three variants that were actually selling, and rebuild the negatives. Over 90 days, with no budget increase, wasted spend fell to 16% and ACOS went from 38.4% to 24.7%. No volume estimate could have flagged that leak. The search term report did.
The one blind spot, and why the other four sources exist
Your search term report only contains searches your ads already entered and got clicked on. A relevant search you never bid on, or bid too low to win, simply never appears. That blind spot is the entire job of the other four sources: they find candidates the report cannot see, and the report then judges them.
So the ranking is a division of labor: four sources nominate, the report votes. The costly mistake is skipping the vote, loading a tool list straight into exact match at full bids as if estimated demand were proof of fit.
How to find keywords for Amazon PPC: my 7-step process
Amazon PPC keyword research works best as a loop: start from the evidence you already own, widen the net with Amazon's first-party data and the search bar, borrow vocabulary and ASINs from competitors, fill gaps with a keyword tool, filter for relevance, then route everything into campaigns where your own ad data decides.
- Mine your own search term report. Download a 60-day bulk file with search term data included and list every search that has produced orders but is not yet an exact match keyword, auto campaign finds included. Those are proven, not candidates; my search term report guide has the promotion bar.
- Check Brand Analytics if you are brand-registered. Its search query data shows searches where your brand appears, with market-wide clicks and purchases beside your brand's share, so you can compare your conversion rate with the market's. Searches that earn you impressions but never reach your search term report are your first gap list.
- Type your root terms into the Amazon search bar. Enter the two or three phrases that name the product, note every suggestion, then add a letter after each root to shake loose more. I do this in a private browser window so my own shopping history stays out of it. Record phrasing, not popularity.
- Read the listings that already win. Search each root and open the top organic results. Their titles and bullets show the vocabulary your niche buys with. The ASINs are targets too: products your offer matches or beats on price and review count belong in a competitor targeting campaign.
- Fill gaps with a keyword tool. Run your roots and a few competitor ASINs through a third-party tool to catch phrasing the first four sources missed and to separate big candidates from tiny ones. Read every volume figure as a directional estimate for sorting, never as a reason to bid higher.
- Filter every candidate for relevance. Search each term on Amazon and look at page one. If the results do not look like your product, cut the term, whatever its estimated volume. Set competitor brand names aside as a separate targeting decision, and send your own brand terms to a brand campaign.
- Route by evidence, then let the report judge. Proven converters go into exact match. Everything else enters a broad and phrase research campaign bid below your fair price per click (my match types guide covers how each behaves). After about one order's worth of clicks, promote the winners, reprice expensive converters, negate the terms that never convert, and go back to step one.
Once the list is ready, the free Campaign Builder turns your SKUs and keywords into a ready-to-upload bulk file, including the negatives that keep each search owned by one campaign.
What if the product is new and has no search term data yet?
A new product with no ad history builds its first keyword list from the search bar, competitor listings, and keyword tools, since it has no search term report and usually little Brand Analytics data. That list is a hypothesis. The first job of the launch campaigns is to produce real search term data fast enough to test it.
At launch I lean on the auto campaign harder than at any other time: its four targeting groups (close and loose match, substitutes, and complements) hunt for searches and competitor ASINs on their own. That is why my benchmark table tolerates wasted spend under 50–60% and auto campaign share at or under 40% in the Awareness phase, tightening to under 15% wasted spend and 15% or less auto share by Profitability. A launch pays tuition on guesses.
Auto targeting also works from your listing, so the vocabulary you gathered in steps three and four belongs in your title and bullets as well. The full launch sequence is in my product launch PPC guide, and how ad sales feed organic rank is in my organic ranking guide.
A worked example: pricing a keyword list before you spend on it
Before adding a list of untested keywords, work out what testing them will cost on your own numbers. One order's worth of clicks is the minimum fair trial for a keyword, so your cost per order is roughly what each candidate costs to judge. A long list of guesses is a real invoice, not a free brainstorm.
Take the sample account behind the worked examples across this site: 45,672 clicks and 4,140 orders, bought with $36,303 of ad spend and worth $111,058 in ad sales. Those totals work out to 32.7% ACOS, 9.1% conversion, and $0.79 per click.
At a 9.1% conversion rate, an order arrives roughly every eleven clicks, and on this account that eleven-click trial costs about $8.77, simply spend divided by orders. That is the price of one fair verdict on one keyword, so a hundred untested candidates carry roughly $877 of tuition.
Now hold the $0.79 click and the average order value steady, and change only conversion rate. A keyword converting at half the account's rate, about 4.5%, runs an ACOS near 65%, well above even the Awareness bar of under 40% on my benchmark table. One converting at double, about 18%, lands near 16%, just outside the Profitability bar of under 15%. Same click, same basket, a fourfold gap in ACOS, and search volume appears in neither result. On the first keyword, high volume would only mean losing money in bigger batches. The free Audit Dashboard demo shows this read on an anonymized real account.
Are Amazon keyword research tools worth paying for?
Third-party keyword tools are worth it for two jobs: generating a broad candidate list fast, and estimating which candidates are large or tiny so you can prioritize. They are not worth trusting for the one decision that costs money, which is what to bid, because they cannot see how your listing converts on any search.
Each vendor builds its own volume estimate, which is why two tools can give one phrase noticeably different numbers, and why I read them only at low resolution: big, medium, tiny. The damage comes from the workflow they invite: export the biggest list, sort by volume, load it all. That is the shortest route I know to a wasted spend problem.
The free tools I built run the other direction, reading your own bulk file instead of estimating the market. The N-Gram Analyzer surfaces words recurring across your converting and losing searches (the method is in my n-gram analysis guide), and the Wasted Spend Finder ranks keywords spending with no orders. Pay for a tool to nominate keywords if it saves you time; the data that judges them is already yours.
Frequently asked questions
What is the best keyword research tool for Amazon PPC?
The best source is one you already have: the search term report in your Amazon Ads account, the only one that pairs each search with what your ads spent on it and the orders that came back. Brand-registered sellers should add Brand Analytics, which is free. Paid third-party tools help with brainstorming and sizing, but their volumes are estimates.
Are keyword tool search volumes accurate for Amazon?
Treat them as directional. Each vendor builds its own estimate, so one phrase can get noticeably different numbers from different tools. They can tell a large phrase from a tiny one; they cannot set a bid. Even a perfect volume figure would leave out the number that decides ACOS: how your listing converts on that search.
Do I need Brand Registry for Amazon keyword research?
Only for Brand Analytics, which gives brand-registered sellers Amazon's own search query data for searches where their brand appears. Everything else works without it: your search term report, auto campaign results, the search bar, competitor listings, and third-party tools. Without Brand Registry you lose a strong supplement, not the core of the method.
How many keywords should I target in Amazon PPC?
As many as your budget can give a fair trial, usually fewer than a tool export suggests. Divide your monthly research budget by your cost per order (ad spend divided by orders). The result is roughly how many unproven keywords can each earn one order's worth of clicks that month. Anything beyond that sits in the account without reaching a verdict.
How often should I redo Amazon keyword research?
The search term report half never stops: I read it monthly as routine optimization, so new keywords arrive as a byproduct of running ads. The other sources I rerun when something changes: a new variant, a new competitor on page one, a seasonal shift in what people type, or discovery campaigns that stop finding anything new.
Sources
- Best practices for your Sponsored Products ads, Amazon Ads
Start with the keyword data you already own
Your most trustworthy keyword source is already inside a bulk file you can download in minutes. Run it through the free Account Health Snapshot (no email, no account, nothing stored) and it grades wasted spend, auto campaign share, and six other metrics against your phase's benchmarks, a fast read on whether your keyword list earns its budget. If it does not, bring the file to the free 30-minute diagnosis call, where I sort through which searches to keep, reprice, cut, or add with you.